ROE, Edinburgh, 20 April Observational Constraints on the Acceleration Discrepancy Problem. Stacy McGaugh University of Maryland

Transcription

1 ROE, Edinburgh, 20 April 2006 Observational Constraints on the Acceleration Discrepancy Problem Stacy McGaugh University of Maryland

2 What gets us into trouble is not what we don t know. It s what we know for sure that just aint so. - Mark Twain i.e., we shouldn t be overly confident that the universe is filled with some new form of invisible, non-baryonic mass (such as WIMPs) until we actually detect the stuff directly in the laboratory. Current cosmology (ΛCDM) invokes not one but two aethers (dark matter and dark energy); let us be careful not to fall into the same conceptual trap that led classical physicists to infer that Maxwell s theory required aether. It is at least conceivable that there could be a theory which captures the successes of cosmology without the excess baggage.

3 The Dark Matter Tree: The roots represent the empirical roots of the problem; the branches the various proposed solutions. Dark matter solutions are represented on the left branch; modifications of dynamical laws are represented on the right branch. This review focusses on the phenomenology of the mass discrepancy, which appears at a particular acceleration scale. By request, I focus on rotation curves, but will also touch on other data. Time restrictions limit discussion of theories to the specific case of MOND. MSTG CWG TeVeS BSTV?

4 V flat dark matter baryons stars gas The rotation curve and mass model of NGC 6946 (pictured on the first slide). The baryons are the sum of stars and gas computed from the observed mass distribution for the case of maximum disk. Dark matter is whatever is left over.

5 Tully-Fisher relation between luminosity/mass and rotation speed In addition to individual mass models, galaxies adhere to strong global relations like Tully-Fisher Bothun et al. (1985) H-band

6 Newton says V 2 = GM/R. Equivalently, Σ = M/R 2 V 4 = G 2 MΣ TF Relation Therefore Different Σ should mean different TF normalization. The Tully-Fisher relation is not well understood in the context of dark matter. There are many hand-waving models, none of which are completely satisfactory. μ = -2.5 logσ +C One expects, from basic physics, that variations in the distribution of baryonic mass should have an impact on the Tully-Fisher relation (lines). They do not (data). See discussions in McGaugh & de Blok 1998, ApJ, 499, 41 Courteau & Rix 1999, ApJ, 513, 561

7 No Residuals from TF rel n Indeed, the residuals from Tully-Fisher are nearly to totally imperceptible, depending weakly on the choice of circular velocity measured. This causes a fine-tuning problem which is generic to any flavor of dark matter...

8 Requires fine balance between dark & baryonic mass McGaugh 2005, Phys. Rev. Lett. 95, The contribution of the baryonic (filled points) and dark matter (open points) to any given point along the rotation curve must be finely balanced, like a see-saw. As the baryonic contribution increases with baryonic surface density, the dark matter contribution decreases. The two components know intimately about each other...

9 Renzo s Rule: When you see a feature in the light, you see a corresponding feature in the rotation curve. (Sancisi 1995, private communication) (See also Sancisi 2004, IAU 220, 233) The distribution of mass is coupled to the distribution of light. Quantify by defining the Mass Discrepancy: D = V 2 V 2 b = V 2 Υ v 2 + V 2 g This is essentially the ratio of total to baryonic mass. It depends on the stellar mass-to-light ratio, Υ, for which we can explore many possibilities.

10 McGaugh 2004, ApJ, 609, 652 The mass discrepancy does not care about a particular length scale. It does correlate strongly with the centripetal acceleration. 74 galaxies > 1000 points (all data) 60 galaxies > 600 points (errors < 5%) radius orbital frequency acceleration

11 Before we explore the choice of stellar mass-to-light ratio, note that the basic phenomenology is already present in the data, without any free parameters. K -band data (Verheijen 1997) 30 galaxies 220 independent points The K -band dynamical mass-to-light ratio correlates with surface brightness. Note that acceleration relates to surface density: g N = GΣ

12 Different choices of Stellar Mass-to-Light Ratio McGaugh 2004, ApJ, 609, 652

13 MOND MOdified Newtonian Dynamics introduced by Moti Milgrom in 1983 instead of dark matter, suppose the force law changes such that for a >> a o, a g N. for a << a o, a (g N a o ) where g N = GM/R 2 is the usual Newtonain acceleration. More generally, these limits are connected by a smooth interpolation fcn (a/a o ) so that (a/a o ) a = g N. MOND can be interpreted as a modification of either inertia (F = ma) or gravity (the Poisson eqn).

14 MOND predictions The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios Disk Galaxies with low surface brightness provide particularly strong tests None of the following data existed in At that time, LSB galaxies were widely thought not to exist.

15 MOND predictions The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness! The lack of residuals from Tully-Fisher an a priori prediction Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios

16 Test TF slope by extrapolation to very low velocities: (McGaugh 2005)

17 Pizagno et al. (2005) ApJ, 633, 844 This is typical of the range covered by most Tully-Fisher studies.

18 McGaugh (2005) ApJ, 632, 859

19 green line: best fit for mass-to-light ratio estimator with minimum scatter: M d = 50V 4 (M ) (km s 1 )

20 The green line is a good predictor of the location of the extreme dwarfs. The slope is steep (4; the nominal prediction of CDM is 3) Mass estimates adopted from the original authors (table). Vertical error bars represent the range from minimum to maximum disk. Horizontal error bars include random uncertainties plus the range of the asymmetric drift correction. These more accurate, independent data confirm the steep slope found by McGaugh et al. (2000) ApJ, 533, L99

21 The Tully-Fisher Relation MOND predictions Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Other a priori MOND predictions Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios

22 MOND predictions The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios

23 mass surface density = V 2 /(Gh) Not a fit MOND predictions The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness No dark matter, so these better correlate Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios surface brightness

24 MOND In essence, MOND is a phenomenological scaling relation that maps the Newtonian baryonic rotation curve to the observed one with only a single free parameter, the stellar mass-to-light ratio.

25 A detailed look at four galaxies. (Sanders & McGaugh 2002)

26 Begeman, Broeils, & Sanders (1991) Begeman (1987): HI data Blais-Ouellette et al. (2004) Hα Fabry-Perot Daigle et al. (2006) Hα Fabry-Perot Excellent data are available for some galaxies, often from independent data sets. The HI generally traces further out; Hα generally has better resolution. In this case, Hα goes silly when only one side traces the rotation. All three data sets trace the kink at 3 kpc, which is reflected in the photometry: recall Renzo s rule.

27 predictive power: zero free parameters MOND fit K -band stellar population prediction Sanders & Verheijen (1998) The K -band is perhaps the best tracer of stellar mass available. One can use stellar population synthesis models (e.g., Bell et al. 2003) to predict the mass-to-light ratio. This does a good job of predicting the mass-to-light ratio needed in MOND fits. In effect, one can predict the rotation curve completely from the photometry.

28 M33 Sanders (1996) Color gradients should correspond to gradients in the mass-to-light ratio.

29 M33 color gradient corrected This example shows the effect of correcting for the observed gradient.

30 Begeman, Broeils, & Sanders (1991) NGC 1560 This case is interesting for the prominent kink at large radii, where the quasi-spherical dark matter halo should dominate. (Recall Renzo s rule.) Υ = 0.97

31 Begeman, Broeils, & Sanders (1991) NGC 1560 Begeman et al. found a better fit if they increased the distance from the value of 3.0 Mpc estimated in 1991 to 3.4 Mpc. The modern D = 3.45 Mpc, as measured by the Tip of the Red Giant Branch method (Karachentsev et al. 2004) Υ = 0.44

32 Every rotation curve provides a strong test of MOND Sanders & McGaugh 2002, ARA&A, 40, 263 which can be applied to many

33 Sanders & McGaugh 2002, ARA&A, 40, 263 many

34 Sanders & McGaugh 2002, ARA&A, 40, 263 many cases.

35 Residuals of MOND fits Taken together, the fits are pretty good. There is a slight hint of non-circular motion at small radii, though this is considerably less than commonly invoked for cuspy halos.

36 MOND predictions The success of detailed rotation curve fits is highly non-trivial. Once the form of the force law is specified, the dynamics must follow from the observed baryon distribution. This procedure is much, much, much more strongly constrained than fits with an invisible dark matter component, which can be arranged however needed. Such fits have a minimum of three free parameters, resulting in enormous freedom and numerous degeneracies. The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios

37 The single free parameter of a MOND fit is the stellar mass-to-light ratio. This is subject to an independent check against the expectations of stellar population synthesis models. These compare favorably (lines from Bell et al. 2003). MOND fit mass-to-light ratios reproduce not only the mean expected value, but also the trend expected with color and the smaller scatter expected in redder bands.

38 MOND predictions The Tully-Fisher Relation Slope = 4 Normalization = 1/(a 0 G) Fundamentally a relation between Disk Mass and V flat No Dependence on Surface Brightness Dependence of conventional M/L on radius and surface brightness Rotation Curve Shapes Surface Density ~ Surface Brightness Detailed Rotation Curve Fits Stellar Population Mass-to-Light Ratios What are we suppose to conclude from this? That MOND is wrong?

39 There is a considerable amount of phenomenological information beyond rotation curves. (McGaugh 1996) GΣ = a 0 For example, the disk stability limit (Milgrom 1989). This provides a natural explanation for the maximum in the surface brightness distribution (e.g., Freeman s Limit). Disks below the critical surface density are stabilized by MOND; those above are in the Newtonian regime and subject to the usual instabilities. This scale has to be inserted by hand into dark matter models (e.g., Dalcanton et al. 1997). The same scale has been found in SDSS dividing disks and ellipticals (e.g., Kauffmann et al. 2004).

40 m=2 growth rate disk stability MOND DM surface density The stability properties of high surface density disks are indistinguishable in dark matter and MOND. However, the two diverge as one goes to low surface density galaxies deep in the MOND regime. The large dark-to-luminous mass ratios in these systems tend to over-stabilize the disks (see also Mihos et al. 1997). Familiar features like bars and spiral arms are readily understood as disk dynamical features, provided the disk selfgravity is important. The presence of such features in LSB galaxies provides another clue... Figure from Brada s Ph.D. thesis (1996). See also Brada & Milgrom (1998).

41 Disk Masses from Density Waves LSB galaxies got spiral arms! McGaugh & de Blok (1998) predicted that, if conventional density wave analysis were applied to constrain the massto-light ratios of LSB disks, one would infer very high mass-to-light ratios. This follows from the need for disk self gravity to drive spiral features. Subsequent analyses (e.g., Fuchs 2002) found exactly this. Galaxy (M/L) * F F F F568-V1 16 UGC UGC UGC ESO ESO ESO ESO from B. Fuchs, astro-ph/ Big (M/L) * s! not LSBs I am familiar with

42 minimum velocity dispersion? For a disk of a given thickness, MOND can support a higher vertical velocity dispersion than a Newtonian disk plus quasi-spherical dark matter halo. This difference is small at high surface brightness, but becomes pronounced as one goes to low surface densities. MOND provides a natural explanation for the minimum ~7 km/s velocity dispersion frequently measured and for very thin LSB disks seen edge-on.

43 Ellipticals Clusters a 0 Globular Clusters Giant Molecular Clouds dwarf spheroidals Many systems fall within a factor of a few of the critical acceleration scale.

44 dwarf spheroidal satellites of the Milky Way Carina Draco Fornax The dwarf spheroidal satellites of the Milky Way are among the lowest acceleration systems known. As you can see, some of these systems are so low surface density that they are hardly even there!

45 M = 2σ2 G eff R V G eff = G µ(x) M = 81 4 σ 4 a 0 G The mass estimator in MOND depends on whether the internal field of the dwarf or the external field of the Milky Way dominates. In 7 of 9 cases MOND gives sensible results. In 2 cases it gives M/L too high. In these cases, it is not clear (at 1 sigma) which estimator should be employed. One can spin the interpretation either way here... which is the forest and which are the trees?

46 Ellipticals Milgrom (1983); Milgrom & Sanders (2003) Giant ellipticals are in the Newtonian regime in their core, so one expects (in MOND) to infer little need for dark matter until large radii.

47 The data for giant ellipticals have only recently become interesting in this context. So far, these data appear consistent with MOND. Romanowsky et al. (2003) Milgrom & Sanders (2003)

48 More generally, one wonders why one continues to see the mass discrepancy appear at a particular acceleration scale when ΛCDM predicts such a vast swath of possibilities. NGC 3379 globular clusters Bergond et al. (2006)

49 Clusters of Galaxies (Sanders & McGaugh 2002) On the scale of individual galaxies, MOND clearly performs better than CDM. The opposite is true in rich clusters of galaxies, where MOND does not suffice to explain the entire mass discrepancy - one needs more mass (neutrinos?).

50 Galaxies and clusters on the same baryonic mass-circular velocity relation. The red line is the prediction of MOND, the orange line that of ΛCDM. Sanders (2003) Reiprich (2001)

51 This is the same as the preceding plot but with a slope of 4 taken out. The difference between the red line and the green triangles is the residual mass discrepancy for clusters in MOND. The orange (ΛCDM) line is more consistent with these data. It does rather worse for galaxies. More generally, ~1000 km/s seems to be a break point in the phenomenology. Above this scale, the universe looks like ΛCDM. Below this scale, it looks like MOND.

52 The 3rd year WMAP data compared to various predictions: pre-boomerang ΛCDM, the fit to the first year WMAP data sans tilt, and no-cdm (McGaugh 1999; 2004). WMAP3 clearly shows power in excess of the low third peak predicted by no-cdm. I thought this would prove fatal to modified gravity theories, but at this conference Skordis & Ferreira showed that I was wrong to think so.

53 BBN: ω b = Ω b h 2 η 10 Measurements of BBN abundances (compiled by McGaugh 2004). The WMAP data without CDM favor a lower baryon density which is more consistent with the bulk of independent data than is the fit with CDM.

54 Sanders (2001); Sanders & McGaugh (2002) see also Nusser (2001); Kneib & Gibson (2002) A common misconception is that non-baryonic cold dark matter is required to grow structure. This is only true in the context of GR. If we consider more general theories, the growth rate can be more rapid, potentially achieving the effect usually attributed to dark matter.

55 X? Disk Stability Freeman limit in surface brightness distribution thin disks velocity dispersions Microwave background 1st:2nd peak amplitude; BBN early reionization enhanced ISW effect? 3rd peak? see Skorids et al X Other MOND tests LSB disks not over-stabilized MOND does well in a variety of tests, not just those with Dwarf Spheroidals? rotation curves. It certainly has problems (e.g., rich clusters; how do galaxies merge?) but it is not obvious that they are worse than those faced by ΛCDM (e.g., not just one but two Giant Ellipticals invisible components; heinous fine-tuning problems to reproduce rotation curve phenomenology). What we need are Clusters of Galaxies rigorous predictions that subject theories to falsification. I am most suspicious of theories that claim to fit everything all the Structure Formation time. We should take especial care not to be too credulous of claims that happen to support our most favored theory.